1. Field of the Invention
The invention is related to the field of voice recognition, and in particular, to a system that provides customized voice recognition data to various types of voice recognition devices.
2. Description of the Prior Art
A device capable of voice recognition receives an audible voice command from a user and translates the voice command into an electronic or optical signal indicating the command. There are two primary types of voice recognition—speaker dependent and speaker independent. With speaker dependent voice recognition, the device recognizes voice commands by using a user profile that is specific to the user. With speaker independent voice recognition, the device recognizes voice commands with a generic profile that is the same for all users. Speaker dependent voice recognition devices generally operate better than speaker independent devices.
For speaker dependent voice recognition, the user creates their user profile in a time-consuming process referred to as “training”. The user trains a device by recording several voice commands prior to normal device operation to generate a user profile with the pre-recorded voice commands. During normal operation, the device then compares newly received voice commands to the pattern made from pre-recorded voice commands to translate the new voice command into an electric signal indicating the command. The device then implements the command. The device may also modify the pre-recorded commands based on successful command translations or based on the introduction of new commands.
Unfortunately, the above system requires the user to re-train each new voice recognition device. To address this problem, centralized voice recognition systems have been developed to house multiple user profiles. When a device receives a voice command from a user, the device transfers an electronic version of the voice command to the central system along with a user identifier (ID). The central system uses the user ID to retrieve the user profile and uses the user profile to translate the voice command. The central system then transfers the translated command to the appropriate device for implementation. Advantageously, the user may use voice commands to control multiple devices without re-training each device. Unfortunately, centralized command translation requires every device to transfer each voice command over a quality communication link to the central system for voice translation.
To address this problem, a central system that stores multiple user profiles may be adapted to download a user's profile to the given device that the user desires to operate. When a user starts to operate the device, the device transfers the user ID to the central system. The central system uses the user ID to retrieve the appropriate user profile and transfer that user profile to the device. The device then uses the selected user profile for voice command translation. Advantageously, the user may operate various devices using their own profile without re-training and without centralized command translation.
Unfortunately, this adapted central system treats all devices the same. Different types of devices may have different voice recognition capabilities and commands. Different devices may also use different protocols for communicating with the central system. For example, a voice-activated telephone will have different capabilities, commands, and protocols than a voice-activated toy or a voice-activated word processor—although some capabilities, commands, and protocols may be the same. A voice recognition system is needed that can provide voice recognition data that is customized for both the user and the device.